panjinquan/tensorflow_models_learning

tensorflow GoogleNet inception V1 V2 V3 V4

tensorflow_models_learning

老铁要是觉得不错，给个“star”

1.生成record训练数据

dataset已经包含了训练和测试的图片，请直接运行create_tf_record.py

对于InceptionNet V1:设置resize_height和resize_width = 224
对于InceptionNet V3:设置resize_height和resize_width = 299
其他模型，请根据输入需要设置resize_height和resize_width的大小

if __name__ == '__main__':
    # 参数设置
    resize_height = 224  # 指定存储图片高度
    resize_width = 224  # 指定存储图片宽度
    shuffle=True
    log=5
    # 产生train.record文件
    image_dir='dataset/train'
    train_labels = 'dataset/train.txt'  # 图片路径
    train_record_output = 'dataset/record/train{}.tfrecords'.format(resize_height)
    create_records(image_dir,train_labels, train_record_output, resize_height, resize_width,shuffle,log)
    train_nums=get_example_nums(train_record_output)
    print("save train example nums={}".format(train_nums))

    # 产生val.record文件
    image_dir='dataset/val'
    val_labels = 'dataset/val.txt'  # 图片路径
    val_record_output = 'dataset/record/val{}.tfrecords'.format(resize_height)
    create_records(image_dir,val_labels, val_record_output, resize_height, resize_width,shuffle,log)
    val_nums=get_example_nums(val_record_output)
    print("save val example nums={}".format(val_nums))

    # 测试显示函数
    # disp_records(train_record_output,resize_height, resize_width)
    batch_test(train_record_output,resize_height, resize_width)

2.训练过程

目前提供VGG、inception_v1、inception_v3、mobilenet_v以及resnet_v1的训练文件，只需要生成tfrecord数据，即可开始训练

训练VGG请直接运行：vgg_train_val.py
训练inception_v1请直接运行：inception_v1_train_val.py
训练inception_v3请直接运行：inception_v3_train_val.py
训练mobilenet_v1请直接运行：mobilenet_train_val.py
其他模型，请参考训练文件进行修改

3.资源下载

• 本项目详细说明，请参考鄙人博客资料：

《使用自己的数据集训练GoogLenet InceptionNet V1 V2 V3模型》: https://panjinquan.blog.csdn.net/article/details/81560537
《tensorflow实现将ckpt转pb文件》: https://panjinquan.blog.csdn.net/article/details/82218092
《使用自己的数据集训练MobileNet、ResNet实现图像分类（TensorFlow）》https://panjinquan.blog.csdn.net/article/details/88252699 预训练模型下载地址: https://download.csdn.net/download/guyuealian/10610847

• 老铁要是觉得不错，给个“star”
• tensorflow-gpu==1.4.0

tensorflow_models_learning's Issues

如何修改输入节点

ckpt转pb的时候，因为所保存的ckpt文件使用batch方式训练，所以所转换的bp模型的最开始的两个节点为FIFOQueueV2和QueueDequeueManyV2, 如何把修改这个pb模型的输入节点呢，我想把输入改为input（image）这样。

请问：直接运行create_tf_record.py报错，该怎么处理

tensorflow.python.framework.errors_impl.NotFoundError: Failed to create a NewWriteableFile: dataset/record/train224.tfrecords : ϵͳ\udcd5Ҳ\udcbb\udcb5\udcbdָ\udcb6\udca8\udcb5\udcc4·\udcbe\udcb6\udca1\udca3
; No such process

建议生成tfrecords时能够分片生成

不然单独一个文件太大了，不方便传输

直接clone项目，会报错，下载不下来

可视化过程

老哥 这个代码怎么加可视化呢 我的邮箱是[email protected] 真心希望老哥能给个解决办法

报错如何修改？

    for lines in lines_list:
        line=lines.rstrip().split(' ')
        label=[]
        for i in range(labels_num):
            label.append(int(line[i+1]))
        images.append(line[0])
        labels.append(label)

File "/home/admin-pc/Documents/tensorflow_models_learning-master/create_tf_record.py", line 77, in load_labels_file
label.append(int(line[i+1]))
ValueError: invalid literal for int() with base 10: ''

输入/输出节点

请问convert_pb文件中的input_image_tensor、input_keep_prob_tensor、input_is_training_tensor、output_tensor_name如何确定，freeze_graph函数中for op 打印出来的没有这几个值，ckpt文件的计算图输出也没有这几个值？

create_tf_record.py

Traceback (most recent call last):
File "/home/z/tensorflow_models_learning-master/create_tf_record.py", line 303, in
train_nums=get_example_nums(train_record_output)
File "/home/z/tensorflow_models_learning-master/create_tf_record.py", line 39, in get_example_nums
for record in tf.python_io.tf_record_iterator(tf_records_filenames):
File "/home/z/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/lib/io/tf_record.py", line 174, in tf_record_iterator
compat.as_bytes(path), 0, compat.as_bytes(compression_type), status)
File "/home/z/anaconda3/envs/tensorflow/lib/python3.6/site-packages/tensorflow/python/framework/errors_impl.py", line 528, in exit
c_api.TF_GetCode(self.status.status))
tensorflow.python.framework.errors_impl.NotFoundError: /home/z/tensorflow_models_learning-master/dataset/record/train224.tfrecords; No such file or directory

博主，总是出现这个错误是怎么了？

为什么loss一直为0，精度一直为1，训练的是inception_v1

2019-11-27 22:04:34.371258: Step [0] train Loss : 0.000000, training accuracy : 1
2019-11-27 22:04:57.633260: Step [0] val Loss : 0.000000, val accuracy : 1
------save:models\best_models_0_1.0000.ckpt
2019-11-27 22:05:36.576254: Step [10] train Loss : 0.000000, training accuracy : 1
2019-11-27 22:06:12.278251: Step [20] train Loss : 0.000000, training accuracy : 1
2019-11-27 22:06:32.546254: Step [20] val Loss : 0.000000, val accuracy : 1
........

单张图片测试时，所有图片预测结果相同

用自己训练好的模型测试，在测试集上单张图片测试结果相同。
拿之前的验证集测试，如果batch size设为1 预测结果同测试集一样，都是以同样的概率预测为同一个类别，但batch size不为1 比如调整为64 模型就预测正常，这是为什么呢

loss

运行训练代码时报错如上，请问应该如何解决

恳请请教，Froze ERROR！

想跟你请教，根据你的方法或代码保存后，我再导入后缀为meta图并赋予ckpt中的参数，是可以正常跑预测的。但是forzen为pb文件时就报‘INFO:tensorflow:Froze 0 variables.INFO:tensorflow:Converted 0 variables to const ops.’，并且，导入pb文件，再运用预测就会报缺少XX之类的错误。恳请指教，谢谢！

分类63个但是精度无法提高 感觉问题比较大

2020-11-02 15:37:57.208234: Step [0] train Loss : 4.451777, training accuracy : 0.1875
2020-11-02 15:38:42.723464: Step [0] val Loss : 4.219480, val accuracy : 0.0174923
2020-11-02 15:38:55.136853: Step [100] train Loss : 3.162268, training accuracy : 0
2020-11-02 15:39:07.425743: Step [200] train Loss : 1.888495, training accuracy : 0
2020-11-02 15:39:53.201844: Step [200] val Loss : 13.010412, val accuracy : 0.0197906
2020-11-02 15:40:05.477731: Step [300] train Loss : 1.710598, training accuracy : 0
2020-11-02 15:40:17.726494: Step [400] train Loss : 2.285333, training accuracy : 0.125
2020-11-02 15:41:03.523871: Step [400] val Loss : 9.567571, val accuracy : 0.0194076
2020-11-02 15:41:15.935632: Step [500] train Loss : 0.689892, training accuracy : 0
2020-11-02 15:41:28.475818: Step [600] train Loss : 0.531735, training accuracy : 0
2020-11-02 15:42:14.895427: Step [600] val Loss : 10.680215, val accuracy : 0.0190245
2020-11-02 15:42:27.410021: Step [700] train Loss : 0.294478, training accuracy : 0
2020-11-02 15:42:39.976662: Step [800] train Loss : 0.587241, training accuracy : 0
2020-11-02 15:43:26.770598: Step [800] val Loss : 7.437372, val accuracy : 0.0190245
2020-11-02 15:43:39.309107: Step [900] train Loss : 0.352322, training accuracy : 0
2020-11-02 15:43:51.995333: Step [1000] train Loss : 0.210653, training accuracy : 0
2020-11-02 15:44:39.661366: Step [1000] val Loss : 7.019403, val accuracy : 0.0190245
2020-11-02 15:44:52.410022: Step [1100] train Loss : 0.054857, training accuracy : 0
2020-11-02 15:45:05.355245: Step [1200] train Loss : 0.053495, training accuracy : 0
2020-11-02 15:45:52.886870: Step [1200] val Loss : 6.476278, val accuracy : 0.0190245
2020-11-02 15:46:05.563014: Step [1300] train Loss : 0.017112, training accuracy : 0.0625
2020-11-02 15:46:18.331558: Step [1400] train Loss : 0.276403, training accuracy : 0
2020-11-02 15:47:05.632001: Step [1400] val Loss : 5.955071, val accuracy : 0.0190245
2020-11-02 15:47:18.504530: Step [1500] train Loss : 0.018878, training accuracy : 0
2020-11-02 15:47:31.372758: Step [1600] train Loss : 0.059408, training accuracy : 0
2020-11-02 15:48:18.951162: Step [1600] val Loss : 5.360252, val accuracy : 0.0190245
2020-11-02 15:48:31.887936: Step [1700] train Loss : 0.028561, training accuracy : 0
2020-11-02 15:48:44.865589: Step [1800] train Loss : 0.008514, training accuracy : 0
2020-11-02 15:49:32.827837: Step [1800] val Loss : 4.431525, val accuracy : 0.0194076
2020-11-02 15:49:45.732202: Step [1900] train Loss : 0.191928, training accuracy : 0
2020-11-02 15:49:59.040962: Step [2000] train Loss : 0.014791, training accuracy : 0.0625
2020-11-02 15:50:46.432237: Step [2000] val Loss : 4.476066, val accuracy : 0.0190245
-----save:models/model.ckpt-2000
2020-11-02 15:51:00.828174: Step [2100] train Loss : 0.010766, training accuracy : 0.0625
2020-11-02 15:51:13.764994: Step [2200] train Loss : 0.085822, training accuracy : 0
2020-11-02 15:52:00.938146: Step [2200] val Loss : 4.355613, val accuracy : 0.0110444
2020-11-02 15:52:13.771588: Step [2300] train Loss : 0.119277, training accuracy : 0
2020-11-02 15:52:26.567043: Step [2400] train Loss : 0.024402, training accuracy : 0
2020-11-02 15:53:14.222747: Step [2400] val Loss : 4.345027, val accuracy : 0.0193437
2020-11-02 15:53:27.150898: Step [2500] train Loss : 0.050773, training accuracy : 0
2020-11-02 15:53:39.925711: Step [2600] train Loss : 0.006547, training accuracy : 0
2020-11-02 15:54:27.511248: Step [2600] val Loss : 4.385560, val accuracy : 0.0111721
2020-11-02 15:54:40.305635: Step [2700] train Loss : 0.011848, training accuracy : 0
2020-11-02 15:54:53.234307: Step [2800] train Loss : 0.012595, training accuracy : 0
2020-11-02 15:55:41.143845: Step [2800] val Loss : 4.392406, val accuracy : 0.0110444
2020-11-02 15:55:53.962262: Step [2900] train Loss : 0.018239, training accuracy : 0
2020-11-02 15:56:06.990641: Step [3000] train Loss : 0.067942, training accuracy : 0
2020-11-02 15:56:54.403528: Step [3000] val Loss : 4.526069, val accuracy : 0.0106614
2020-11-02 15:57:07.287715: Step [3100] train Loss : 0.002908, training accuracy : 0
2020-11-02 15:57:20.071291: Step [3200] train Loss : 0.015377, training accuracy : 0
2020-11-02 15:58:07.453693: Step [3200] val Loss : 4.584523, val accuracy : 0.0180031
2020-11-02 15:58:20.484154: Step [3300] train Loss : 0.003488, training accuracy : 0
2020-11-02 15:58:33.126986: Step [3400] train Loss : 0.077828, training accuracy : 0
2020-11-02 15:59:20.953589: Step [3400] val Loss : 4.706525, val accuracy : 0.0118105
2020-11-02 15:59:33.913899: Step [3500] train Loss : 0.002673, training accuracy : 0.0625
2020-11-02 15:59:46.716906: Step [3600] train Loss : 0.086050, training accuracy : 0
2020-11-02 16:00:34.343889: Step [3600] val Loss : 4.733990, val accuracy : 0.0178115
2020-11-02 16:00:47.121723: Step [3700] train Loss : 0.000803, training accuracy : 0
2020-11-02 16:01:00.026031: Step [3800] train Loss : 0.004648, training accuracy : 0.0625
2020-11-02 16:01:47.546183: Step [3800] val Loss : 4.868133, val accuracy : 0.0179392
2020-11-02 16:02:00.286305: Step [3900] train Loss : 0.002961, training accuracy : 0.125
2020-11-02 16:02:13.520870: Step [4000] train Loss : 0.083561, training accuracy : 0.0625
2020-11-02 16:03:00.840206: Step [4000] val Loss : 5.280580, val accuracy : 0.0179392
-----save:models/model.ckpt-4000
2020-11-02 16:03:14.624543: Step [4100] train Loss : 0.028515, training accuracy : 0
2020-11-02 16:03:27.654577: Step [4200] train Loss : 0.083428, training accuracy : 0
2020-11-02 16:04:15.113641: Step [4200] val Loss : 5.515779, val accuracy : 0.0179392
2020-11-02 16:04:28.035244: Step [4300] train Loss : 0.006602, training accuracy : 0.0625
2020-11-02 16:04:41.012877: Step [4400] train Loss : 0.005454, training accuracy : 0
2020-11-02 16:05:28.710228: Step [4400] val Loss : 4.986513, val accuracy : 0.0162794
2020-11-02 16:05:41.649157: Step [4500] train Loss : 0.117655, training accuracy : 0
2020-11-02 16:05:54.546338: Step [4600] train Loss : 0.008130, training accuracy : 0
2020-11-02 16:06:43.125019: Step [4600] val Loss : 5.318214, val accuracy : 0.0138534
2020-11-02 16:06:56.144484: Step [4700] train Loss : 0.198717, training accuracy : 0
2020-11-02 16:07:09.052229: Step [4800] train Loss : 0.166929, training accuracy : 0
2020-11-02 16:07:56.452739: Step [4800] val Loss : 5.759774, val accuracy : 0.0183861
2020-11-02 16:08:09.284685: Step [4900] train Loss : 0.001290, training accuracy : 0.0625
2020-11-02 16:08:22.219911: Step [5000] train Loss : 0.037633, training accuracy : 0.0625
2020-11-02 16:09:10.040866: Step [5000] val Loss : 6.110466, val accuracy : 0.0179392
2020-11-02 16:09:22.748253: Step [5100] train Loss : 0.047009, training accuracy : 0.0625
2020-11-02 16:09:35.874876: Step [5200] train Loss : 0.107128, training accuracy : 0
2020-11-02 16:10:23.152308: Step [5200] val Loss : 6.624900, val accuracy : 0.0179392
2020-11-02 16:10:36.051775: Step [5300] train Loss : 0.085515, training accuracy : 0.0625
2020-11-02 16:10:48.866156: Step [5400] train Loss : 0.409375, training accuracy : 0
2020-11-02 16:11:36.761074: Step [5400] val Loss : 6.839243, val accuracy : 0.0263023
2020-11-02 16:11:49.638604: Step [5500] train Loss : 0.074292, training accuracy : 0.0625
2020-11-02 16:12:02.772029: Step [5600] train Loss : 0.015727, training accuracy : 0
2020-11-02 16:12:50.155442: Step [5600] val Loss : 6.886412, val accuracy : 0.0256639
2020-11-02 16:13:02.860564: Step [5700] train Loss : 0.011184, training accuracy : 0
2020-11-02 16:13:15.807399: Step [5800] train Loss : 0.105041, training accuracy : 0.0625
2020-11-02 16:14:02.974487: Step [5800] val Loss : 6.774150, val accuracy : 0.0195352
2020-11-02 16:14:15.964116: Step [5900] train Loss : 0.006180, training accuracy : 0
2020-11-02 16:14:28.860469: Step [6000] train Loss : 0.002312, training accuracy : 0
2020-11-02 16:15:16.336347: Step [6000] val Loss : 6.687222, val accuracy : 0.0291113
-----save:models/model.ckpt-6000
2020-11-02 16:15:30.259542: Step [6100] train Loss : 0.000506, training accuracy : 0
2020-11-02 16:15:42.874291: Step [6200] train Loss : 0.002289, training accuracy : 0.0625
2020-11-02 16:16:30.261237: Step [6200] val Loss : 7.348487, val accuracy : 0.0201736
2020-11-02 16:16:43.067244: Step [6300] train Loss : 0.007106, training accuracy : 0.125
2020-11-02 16:16:56.013800: Step [6400] train Loss : 0.000544, training accuracy : 0
2020-11-02 16:17:43.874414: Step [6400] val Loss : 7.402358, val accuracy : 0.0302605
2020-11-02 16:17:56.692698: Step [6500] train Loss : 0.012519, training accuracy : 0
2020-11-02 16:18:09.608623: Step [6600] train Loss : 0.020444, training accuracy : 0.0625
2020-11-02 16:18:57.944717: Step [6600] val Loss : 7.957761, val accuracy : 0.032048
2020-11-02 16:19:10.604547: Step [6700] train Loss : 0.057993, training accuracy : 0.125
2020-11-02 16:19:23.401778: Step [6800] train Loss : 0.062361, training accuracy : 0
2020-11-02 16:20:10.866427: Step [6800] val Loss : 6.631777, val accuracy : 0.0329418
2020-11-02 16:20:23.524615: Step [6900] train Loss : 0.028056, training accuracy : 0
2020-11-02 16:20:36.328313: Step [7000] train Loss : 0.002315, training accuracy : 0
2020-11-02 16:21:23.668482: Step [7000] val Loss : 10.059795, val accuracy : 0.0365807
2020-11-02 16:21:36.660421: Step [7100] train Loss : 0.001399, training accuracy : 0.0625
2020-11-02 16:21:50.029407: Step [7200] train Loss : 0.044736, training accuracy : 0.0625
2020-11-02 16:22:37.962835: Step [7200] val Loss : 9.027893, val accuracy : 0.036453
2020-11-02 16:22:50.644555: Step [7300] train Loss : 0.002088, training accuracy : 0
2020-11-02 16:23:03.500411: Step [7400] train Loss : 0.001914, training accuracy : 0
2020-11-02 16:23:50.540379: Step [7400] val Loss : 9.877556, val accuracy : 0.0377937
2020-11-02 16:24:03.278654: Step [7500] train Loss : 0.010749, training accuracy : 0.0625
2020-11-02 16:24:15.917410: Step [7600] train Loss : 0.000708, training accuracy : 0
2020-11-02 16:25:03.002879: Step [7600] val Loss : 6.635362, val accuracy : 0.0444969
2020-11-02 16:25:15.709085: Step [7700] train Loss : 0.001703, training accuracy : 0
2020-11-02 16:25:28.378377: Step [7800] train Loss : 0.000629, training accuracy : 0.0625
2020-11-02 16:26:15.486690: Step [7800] val Loss : 6.803297, val accuracy : 0.0425179
2020-11-02 16:26:28.027219: Step [7900] train Loss : 0.061680, training accuracy : 0
2020-11-02 16:26:40.677709: Step [8000] train Loss : 0.002516, training accuracy : 0
2020-11-02 16:27:27.783940: Step [8000] val Loss : 6.278635, val accuracy : 0.0385598
-----save:models/model.ckpt-8000
2020-11-02 16:27:41.331630: Step [8100] train Loss : 0.001402, training accuracy : 0.0625
2020-11-02 16:27:54.020844: Step [8200] train Loss : 0.001184, training accuracy : 0.1875
2020-11-02 16:28:41.079053: Step [8200] val Loss : 4.479759, val accuracy : 0.0701609
2020-11-02 16:28:53.780333: Step [8300] train Loss : 0.000839, training accuracy : 0
2020-11-02 16:29:06.536083: Step [8400] train Loss : 0.001597, training accuracy : 0.0625
2020-11-02 16:29:53.773071: Step [8400] val Loss : 4.883051, val accuracy : 0.0725868
2020-11-02 16:30:06.546049: Step [8500] train Loss : 0.013211, training accuracy : 0.0625
2020-11-02 16:30:19.321398: Step [8600] train Loss : 0.004649, training accuracy : 0
2020-11-02 16:31:06.456382: Step [8600] val Loss : 8.681377, val accuracy : 0.0537538
2020-11-02 16:31:19.036534: Step [8700] train Loss : 0.006653, training accuracy : 0.125
2020-11-02 16:31:31.854400: Step [8800] train Loss : 0.002683, training accuracy : 0
2020-11-02 16:32:18.938620: Step [8800] val Loss : 4.728459, val accuracy : 0.104635
2020-11-02 16:32:31.707159: Step [8900] train Loss : 0.020025, training accuracy : 0.125
2020-11-02 16:32:44.462050: Step [9000] train Loss : 0.008877, training accuracy : 0.0625
2020-11-02 16:33:31.752998: Step [9000] val Loss : 3.812795, val accuracy : 0.199694
2020-11-02 16:33:44.462998: Step [9100] train Loss : 0.001572, training accuracy : 0.1875
2020-11-02 16:33:57.337466: Step [9200] train Loss : 0.005766, training accuracy : 0.125
2020-11-02 16:34:44.460697: Step [9200] val Loss : 3.064202, val accuracy : 0.234359
2020-11-02 16:34:57.289857: Step [9300] train Loss : 0.001191, training accuracy : 0
2020-11-02 16:35:09.949805: Step [9400] train Loss : 0.003511, training accuracy : 0.1875
2020-11-02 16:35:57.668908: Step [9400] val Loss : 2.668586, val accuracy : 0.291433
2020-11-02 16:36:10.336091: Step [9500] train Loss : 0.030213, training accuracy : 0.625
2020-11-02 16:36:23.274206: Step [9600] train Loss : 0.000295, training accuracy : 0.875
2020-11-02 16:37:11.567725: Step [9600] val Loss : 1.537270, val accuracy : 0.569523
2020-11-02 16:37:24.354095: Step [9700] train Loss : 0.000855, training accuracy : 0.5
2020-11-02 16:37:37.440869: Step [9800] train Loss : 0.000576, training accuracy : 0.625
2020-11-02 16:38:24.862105: Step [9800] val Loss : 1.792353, val accuracy : 0.449694
2020-11-02 16:38:37.705421: Step [9900] train Loss : 0.137523, training accuracy : 0.625
2020-11-02 16:38:51.008138: Step [10000] train Loss : 0.000887, training accuracy : 0.8125
2020-11-02 16:39:38.602564: Step [10000] val Loss : 1.078526, val accuracy : 0.668156
-----save:models/model.ckpt-10000

UnboundLocalError: local variable 'labels' referenced before assignment

您好，在运行您的create_labels_files.py程序时报错
UnboundLocalError: local variable 'labels' referenced before assignment
在加入全局变量之后报错提示labels未定义，请问这个该如何定义？我定义初始值为0生成的val.txt是个空文档。

预测测试集样本所属的类别，经常发生改变？

用图片训练好inception_v3模型后，再用其他图片做预测，为什么对于测试集样本，每次预测所属的类别经常变动？

如何训练mobilenet——v2模型

修改代码直接调用会有bug
修改如下：
with tf.contrib.slim.arg_scope(mobilenet_v2.training_scope()):
out, endpoints = mobilenet_v2.mobilenet(input_tensor=input_images)

BUG如下：
Traceback (most recent call last):
File "train_v2.py", line 213, in
snapshot_prefix=snapshot_prefix)
File "train_v2.py", line 158, in train
tf.losses.softmax_cross_entropy(onehot_labels=input_labels, logits=out)
File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/ops/losses/losses_impl.py", line 732, in softmax_cross_entropy
logits.get_shape().assert_is_compatible_with(onehot_labels.get_shape())
File "/opt/conda/lib/python3.6/site-packages/tensorflow/python/framework/tensor_shape.py", line 764, in assert_is_compatible_with
raise ValueError("Shapes %s and %s are incompatible" % (self, other))
ValueError: Shapes (?, 1001) and (?, 4) are incompatible

老哥给个模型吧

我只想用googleNet V3模型做一个特征提取，我的数据也没有标签，所以不能训练，请老哥给个模型下载链接，我的邮箱是[email protected]，多谢了

冒昧问一句，请问有可视化的接口吗

我尝试用mean_acc一直没有成功

您好，用您预测的代码加载不了训练好的模型，查了好久也查不出来

NotFoundError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a Variable name or other graph key that is missing from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint. Original error:

Key InceptionV3/AuxLogits/Conv2d_1b_1x1/BatchNorm/beta not found in checkpoint
[[Node: save_1/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_INT64], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save_1/Const_0_0, save_1/RestoreV2/tensor_names, save_1/RestoreV2/shape_and_slices)]]
[[Node: save_1/RestoreV2/_307 = _Recvclient_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_312_save_1/RestoreV2", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]]

老哥有没有其他的方式可以下载你训练好的模型？

Performance issue in /slim/datasets (by P3)

Hello! I've found a performance issue in download_and_convert_cifar10.py: with tf.Session('') as sess(here) is defined in the function _add_to_tfrecord(here) which is repeatedly called in the loop for i in range(_NUM_TRAIN_FILES)(here).

tf.Session being defined repeatedly could lead to incremental overhead. If you define tf.Session out of the loop and pass tf.Session as a parameter to the loop, your program would be much more efficient. Here is the Stack Overflow post to support it.

Looking forward to your reply. Btw, I am very glad to create a PR to fix it if you are too busy.

如何在tf2.0下运行inception_v3_train_val.py

修改inception_v3_train_val.py 增加
import tensorflow.compat.v1 as tf
tf.disable_v2_behavior()
试着在tf2.0下运行，提示错误
import slim.nets.inception_v3 as inception_v3
from nets import inception_utils
AttributeError: module 'tensorflow' has no attribute 'contrib'
contrib已经在2.0中移除，这三个库怎么导入？
多谢！

梯度可视化

ERROR! Session/line number was not unique in database. History logging moved to new session

您好，我運行程式的時候都會遇到這個報錯資訊
ERROR! Session/line number was not unique in database. History logging moved to new session 243
如果我重新啟動程式在運行的話最後數字的地方會累加 請問該如何解決,謝謝

您好，用您预测的代码加载不了训练好的模型，查了好久也查不出来

NotFoundError (see above for traceback): Restoring from checkpoint failed. This is most likely due to a Variable name or other graph key that is missing from the checkpoint. Please ensure that you have not altered the graph expected based on the checkpoint. Original error:

Key InceptionV3/AuxLogits/Conv2d_1b_1x1/BatchNorm/beta not found in checkpoint
[[Node: save_1/RestoreV2 = RestoreV2[dtypes=[DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, ..., DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_FLOAT, DT_INT64], _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_save_1/Const_0_0, save_1/RestoreV2/tensor_names, save_1/RestoreV2/shape_and_slices)]]
[[Node: save_1/RestoreV2/_307 = _Recvclient_terminated=false, recv_device="/job:localhost/replica:0/task:0/device:GPU:0", send_device="/job:localhost/replica:0/task:0/device:CPU:0", send_device_incarnation=1, tensor_name="edge_312_save_1/RestoreV2", tensor_type=DT_FLOAT, _device="/job:localhost/replica:0/task:0/device:GPU:0"]]

预测脚本报错

分类数量为1时，报错too many indices for array

准确率升不上去

您好，我在训练你给数据训练模型的时候，准确率一直都20%作用，我训练了20万次，准确率上不去。你问这是怎么回事啊？

import slim.nets.inception_v3 as inception_v3出现错误

老哥你好，我的tensorflow版本是GPU1.9，运行python inception_v3_train_val.py时，出现如下错误：
import slim.nets.inception_v3 as inception_v3时报错No module named 'nets'
请问老哥怎么解决？ @PanJinquan

训练出来的ckpt文件转pb文件

,mobilenet节点具体名称我要用哪个,可以说一下吗,谢谢了,还有这样训练的模型opencv可以调用吗,谢谢作者.

slim里面是啥？

哈喽，大佬你好

我看到slim，不知道是干啥的，我记得tf中有slim啊，不一回事？